Sentiment Analysis is a core Revinate feature that will help you quickly and easily assess the guest sentiment attributed to your hotel, competitors' hotels, or any service, function, or amenity. It is an automated method for identifying mentions of specific topics of interest (such as "concierge" or "pool") and determining whether the author of the feedback felt positive or negative about that topic. This type of analysis is obviously very powerful, as it allows Revinate to unlock much more value from the unstructured reviews left on public websites.
You can check out our FAQs around Sentiment Analysis here. as well as best practices and suggested uses for Sentiment Analysis and/or view a recording of our learning webinar on Sentiment Analysis.
We've tailored our Sentiment Analysis feature specifically for the hospitality industry. We have identified hundreds of the most important topics for hotels and categorized them based on traditional operations and guest feedback categories ("Facilities," "Staff & Service," "Rooms," "COVID-19" etc.) So you can measure not only how guests feel about your "buffet," but also the overall performance of your "Food & Beverage" operation.
Languages
Sentiment scoring is native to 9 languages:
- English
- Spanish
- German
- French
- Simplified Chinese
- Italian
- Dutch
- Japanese
- Portuguese
Sentiment is scored natively in the original language of the review and rolled up into the overall score for each topic or category. We aim to present all of this very rich data as intuitively as possible so that it's immediately understandable by everyone.
Topics
Sentiment Analysis displays one value per topic triggered in a review. So a review may trigger 5 topics and return only 1 value, even if the topic was mentioned multiple times throughout the review. Revinate refers to this as "review level scoring", because it's an overall score per category across the whole review.
For example, if a guest submits a review and mentions the pool, the gym, and the elevators, all of this would roll up into the facilities category.
On the topics page you can view overall scores for topics and categories. Use the filters to narrow down the data you would like to see. The graph gives you a visual of your scores and mentions of your selected category over time. The table below the graph breaks down information by topics within the category. It shows you the topic's overall score, total mentions, the percent of positive, negative, and neutral reviews about this topic, if scores are trending up or down, and a bar chart giving a visual of the distribution of the general sentiment in the reviews for the topic. Click on the topic or the data in the table to view reviews that correspond with the topic and/or data.
Revinate uses ranges based on Semantria's best practices as the text analytics industry experts to classify a score as positive vs. neutral vs. negative.
Trigger and sentiment words
This software examines each review to determine which categories are triggered based on keywords (e.g., "bed"), as well as their scores based on the presence of other sentiment-bearing words in the same sentence (e.g., "comfortable").
The sentiment software uses advanced linguistic analysis to determine this score. When the software detects words like "extremely", they are processed as multipliers, resulting in a more positive or more negative score depending on the remainder of the phrase or sentence.
Example
The phrase "extremely impressive" would register as a strongly positive sentiment, while "extremely poor" would register as strongly negative.
We display an overall score for each topic that considers all relevant references within the same review. Each category's sentiment score will be on a scale of 0 - 100 and can be interpreted as follows:
Very Positive: 78 - 100
Positive: 56 - 77
Neutral: 49 - 56
Negative: 24 - 48
Very Negative: 0 - 23
This scoring system will provide a simple and easy way for you to understand what elements of your hotel are performing well and what operations still need improvement.
Editing Sentiment Analysis
If you feel that the sentiment of a review or survey has been misinterpreted, you can go in and edit it. To edit a reviews sentiment click into the reviews tab in sentiment analysis.
In the review you are trying to change the sentiment for, click the edit sentiment button.
In the pop-up box, you can add sentiment for subjects that may have been missed by typing in the "Add Sentiment" box and selecting from the drop-down.
You can drag subjects to different levels of sentiment to adjust the reviews overall sentiment.
Sentiment analysis for surveys
For Surveys, you can choose from the drop-down menu of filtering options to choose a question or questions on your survey that you would want to analyze for sentiment. It is also possible to filter for the Likelihood to Recommend question and dive deeper into feedback from your Promoters or Detractors. This feature will analyze the content of open text questions as of August 16th, 2018 but not prior.
Accuracy
It's important to understand while our Sentiment Analysis functionality is highly sophisticated and very powerful, some sentiment scores will not always appear to be accurate. As you can imagine, if you gathered 10 people in a room to interpret certain reviews, you would likely get a variety of answers - especially if the reviews contain poor grammar, misspelled words, or slang.
To ensure the highest possible accuracy, we have partnered with Semantria, an industry leader in text analytics. We are continuously striving to improve our results. However, due to the inherently subjective nature of sentiment and because reviews are unstructured, sentiment scoring will never appear to be completely accurate.
It helps to keep this in mind so that you understand that our Sentiment Analysis feature provides very valuable and "actionable" aggregate scores. For example, while one specific mention of "sheets" might appear to be misclassified as negative, if 89% of guests feel positive about your sheets, you can rest assured that you can focus on other areas of improvement!
Multiple categorizations
Some topics may be included in multiple categories, so while some topics are in more than one category, their mention counts will not be double- counted in the overall mention count. This is why you could see the overall mention count smaller than the sum of the mention count from the other six categories.
Example topics that fall under more than one category:
- "Room cleanliness" falls under both the Rooms and the "Cleanliness" categories
- "Public odor" falls under both the "Cleanliness" and "Facilities" categories
- "Internet access" falls under both the "Rooms" and "Facilities" categories
View our Sentiment Analysis Webinar recording here.
COVID-19 Sentiment Analysis
We have added a new topic for COVID-19, which you can choose under "Facilities -> COVID-19".
The COVID-19 category will track common word references related to COVID (e.g. COVID-19, Corona Virus, etc.).